Lightweight C&C based botnet detection using Aho-Corasick NFA

Abstract

Botnet distinguishes itself from the previous malware by having the characteristics of a C&C channel, using which a Botmaster can control the constituents of the botnet. Even though protocols like IRC, HTTP and DNS are exploited to incorporate C&C channels, previous analysis have shown that the majority of the botnets are usually based on IRC. Consequently in this paper the Aho-Corasick NFA based detection is proposed to detect the C&C instructions which is exchanged in IRC run botnets. However the ability to detect botnet is limited to the existing bot commands. Therefore a counting process which analyses every IRC messages is introduced to detect the existence of malicious codes. This detection method and various existing methods have been evaluated using real-world network traces. The results show that the proposed C&C Instruction based IRC detection method can detect real-world botnets with high accuracy.